JP2008530594A - Method and apparatus for distinguishing buildings from vegetation for terrain modeling - Google Patents

Abstract

A computer-implemented method for processing a digital elevation model (DEM) that contains data about multiple objects. The method determines a circumference-to-area parameter for each object in the DEM (block 502) and compares the circumference-to-area parameter for each object with a threshold to determine whether each object in the DEM is a building 54 or vegetation 56. Including identifying what is.

Description

  Geographic area terrain models can be used for many applications, including flight simulators and floodplain analysis. Furthermore, terrain models of artificial structures (eg, cities) can be extremely useful in applications such as mobile phone antenna placement, city planning, disaster preparedness and analysis, and mapping.

  Various types and methods are currently used to create terrain models. One common terrain model is the digital elevation model (DEM). A DEM is a sampled matrix representation of a geographic area and can be generated in a computer automated manner. In DEM, a coordinate value is associated with a height value. DEMs are typically used to model land where the transition between different elevations (eg valleys, mountains, etc.) is generally smooth from one elevation to the next. In other words, DEM typically models the land as multiple curved surfaces that are “justified” even if there are discontinuities between them. For this reason, DEMs are generally not suitable for modeling artificial structures such as city center skyscrapers with sufficient accuracy for many of the above applications.

  US Pat. No. 6,654,690 to Rahmes et al. Discloses a significant advance in geomorphology. The '690 patent discloses an automated method for creating a terrain model of an area containing land and buildings above it based on randomly spaced data of elevation versus location. The '690 patent is assigned to the assignee of the present invention and is hereby incorporated by reference in its entirety. The method processes the randomly spaced data to generate elevation-positioned gridded data that fits a predetermined position grid, processes the gridded data to distinguish building data from land data, Performing polygon extraction on the data to create a terrain model that includes the land in the region and the buildings above it.

  In particular, a DEM is generated only for land and a DEM is generated only for buildings. Once the building is distinguished from the land, polygon extraction is performed on the building data. The '690 patent creates a terrain model of the area containing land and buildings above it in a relatively quick manner and with improved accuracy. Nevertheless, what is commonly seen is that many vegetation, especially trees, can be treated as buildings. That is, polygon extraction is also performed on data representing trees. This results in a larger number of polygons being used to model the tree than the number of polygons used to model the building. When a terrain model is displayed on the viewer, the modeled vegetation does not look very realistic. As a result, the modeled vegetation is manually removed and replaced with a more realistic model. This is relatively time consuming and labor intensive.

  In view of the above background, it is an object of the present invention to provide a computer-implemented method for distinguishing vegetation from buildings within a digital elevation model (DEM).

  This and other objects, features, and effects according to the present invention are provided by a computer-implemented method for processing a DEM that includes data about multiple objects. The method determines a circumference-to-area parameter for each object in the DEM and compares the circumference-to-area parameter for each object with a threshold to identify whether each object in the DEM is a building or a vegetation Can be included.

  The data for each object includes a height value, and the computer-implemented method further compares the height value for each object identified as vegetation with a height threshold, and the associated height value is high. Re-identifying each vegetation as a building if greater than the threshold may be included.

  In addition, the computer-implemented method further determines a second perimeter parameter for each object identified as a building, compares each second perimeter parameter to a second threshold, Re-identifying each building as a vegetation if the perimeter parameter of is greater than a second threshold.

  The objects identified as buildings can then be separated into building DEMs, and the objects identified as vegetation can be separated into vegetation DEMs. Separate building and vegetation DEMs advantageously allow a more realistic terrain model to be generated with significantly less user intervention.

  The computer-implemented method may further include modeling each building in the building DEM as a vector. Here, each vector may include a plurality of polygons. The computer-implemented method may further include modeling each vegetation in the vegetation DEM as a 3D point.

  Another aspect in accordance with the invention is directed to a computer readable medium having stored thereon a data structure for processing a digital elevation model (DEM) that includes data for a plurality of objects. The computer-readable medium includes a first data field that includes data for determining a circumference-to-area parameter for each object in the DEM, and compares the circumference-to-area parameter for each object with a threshold value for each object in the DEM. And a second data field containing data for identifying whether the object is a building or a vegetation.

  Yet another aspect of the invention is directed to a computer system for terrain modeling. The computer system may have a processor for processing the computer readable medium defined above. A display may be coupled to the processor for displaying the terrain model based on the processing.

  The present invention will now be described more fully with reference to the accompanying drawings. In so doing, preferred embodiments of the invention are shown, but the invention can be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Like reference numerals refer to like elements throughout.

  A collector 50 for collecting terrain data and a system 60 for generating a digital elevation model (DEM) from the collected terrain data will now be described with reference to FIG. The DEM is of an area that includes land 52 and objects on that land, where the objects can be buildings 54 and vegetation 56. Vegetation 56 primarily includes trees, and therefore vegetation and trees are interchangeable. The modeling of land 52, building 54 and vegetation 56 is based on elevation and position data at random and arbitrarily spaced in the area.

  A collector 50 such as light detection and ranging (LIDAR) can be used to collect the randomly spaced data. The randomly spaced data can be nominally a set of non-uniformly spaced positions and altitude measurements. LIDAR collector 50 may be carried by aircraft 70 over an area of interest such as a city. The area may also include features that are relatively small compared to buildings 54 and trees 56, such as roads 58.

  One skilled in the art will recognize that a LIDAR source provides data including elevation versus location information from a single image. In order to provide altitude pair position data, generally, a plurality of optical images of the area taken from various viewpoints are required, but the same information can be obtained from a single LIDAR image. Of course, as those skilled in the art will appreciate, in addition to the LIDAR collector, the present invention uses elevation versus position data from a variety of sources, such as optical (eg, photographic), electro-optic and infrared sources. be able to. The location information provided by the LIDAR data can include, for example, latitude and longitude information, but other suitable location indicators can also be used.

  Once randomly spaced data is collected, the data can be stored on a storage medium 80, such as a magnetic or optical disk, for transfer to the computer 62. Of course, other suitable methods for data transfer may be used as will be readily appreciated by those skilled in the art. The randomly spaced data is used by the computer 62 to generate a DEM for viewing.

  A display 64 is connected to the computer 62 for viewing the DEM. Input devices such as a keyboard 66 and a mouse 68 are also connected to the computer 62. In accordance with the present invention, the computer 62 improves the DEM by generating 1) land only DEM and object only DEM, and then removing noise from the object only DEM, and 2) building the object only DEM and the vegetation DEM. Including a processor 69 for separation.

  The creation and enhancement of the original or initial DEM will now be discussed with reference to the flowchart of FIG. 2 and the computer display screens shown in FIGS. In the flow chart, steps (1) to (19) are first discussed. Here, steps (2)-(19) are considered part of the batch process, as will be discussed in more detail later. In a batch process, some of the blocks shown in the flowchart are discussed more than once. This is because each of these functions is repeated based on a loop process. To better illustrate the method of improving the original DEM, the number of each step discussed is given in parentheses within the corresponding block.

  From the start (block 100), the initial step (1) uses the computer 62 at block 102 to generate the original DEM from randomly spaced data provided via the storage medium 80. Referring to the initial computer screen 200 shown in FIG. 3, the user selects “Generate DEM from points” in field 202. Then, as shown in FIG. 4, a “DEM setting from point” computer screen 204 is generated.

  On the “Set DEM from Point” computer screen 204, the name of the file storing the collected data is entered in the field 206. A point format is selected in field 208. In this case, the points are based on a Universal Transverse Mercator (UTM) grid. The point measurement unit is selected in field 210, for example, meters. The UTM grid contains 60 north-south zones, each zone being 6 degrees wide at latitude. The UTM zone of interest is selected in field 212. For example, zone 15 is selected in field 212. The data resolution is selected in field 214 and the window filter size is selected in field 216. This will be readily recognized by those skilled in the art. The generated original DEM is given to the computer screen 300 as shown in FIG.

  Steps (2)-(19) for improving the original DEM are initiated by selecting “Run Batch Process” in field 220 from the initial computer screen 200 shown in FIG. When a batch process is executed, fields 222, 224 and 226 allow the user to set certain parameters related to the batch process. These parameters are described below.

  Step (2) at block 104 is resampling of the original DEM. Settings related to resampling are provided on the computer screen 230 as shown in FIG. The original DEM resampling has a resolution of, for example, 1 meter, where it is resampled at a lower resolution. Its resolution is set in field 232, for example 30 meters. A window filter size is also selected in field 234. The result is given on a computer screen 302 as shown in FIG. This result is obtained by leveling the objects 54 and 56 on the land 52.

  At block 106, null filling is performed on the resampled DEM. The null fill is associated with the null operation given in field 226 from the initial computer screen 200 shown in FIG. The null operation can be divided into a null expansion or a null fill as provided on the computer screen 240 of FIG. Field 242 corresponds to null extension and field 244 corresponds to null fill. Since null filling is executed, a computer screen 250 as shown in FIG. 9 is displayed. Settings related to null fill include a field 252 for the fill method, a field 254 for the number of fill passes to be performed, and a field 256 for the fill reach. The resulting resampled DEM after null filling is provided on the computer screen 304 as shown in FIG.

  At block 108, DEM subtraction is performed. The computer screen 260 is related to the DEM subtraction shown in FIG. The threshold used in the DEM subtraction is selected in field 262. The resampled DEM after null filling in step (3) is subtracted from the original DEM in step (1) to generate an object only DEM. An object only DEM is provided on a computer screen 306 as shown in FIG.

  At block 110, step (5) is another DEM subtraction. Only the object DEM from step (4) is removed from the original DEM in step (1) to generate a DEM without the object. This DEM is given on a computer screen 308 as shown in FIG.

  In block 112, corresponding to step (6), a null extension is performed on the DEM without the object. As shown in FIG. 14, the computer screen 270 is associated with a null extension. Null is expanded corresponding to the value selected in field 272. The null extension ensures that all objects are removed and given on the computer screen 310 and the result is a DEM with no objects at 1 meter resolution as shown in FIG.

  The execution batch process loops back to block 104 for step (7). Block 104 performs resampling on the objectless DEM as given in FIG. This resampling is performed at a lower resolution, i.e. a resolution of 1 to 30 meters. The result is given on a computer screen 312 as shown in FIG.

  At block 106, a second null fill is performed corresponding to step (8). This second null fill is performed on the DEM without the resampled object as given in block 112. This process generates a land only DEM as given on the computer screen 314 of FIG. At block 108, a second DEM subtraction is performed corresponding to step (9). The land only DEM from step (8) is now subtracted from the original DEM in step (1) to generate only the enhanced object DEM. The enhanced object only DEM is provided on a computer screen 316 as shown in FIG. In block 110, step (10) is another DEM removal step. Only the enhanced object from step (9) is removed from the original DEM in step (1) to produce an enhanced, object-free DEM. This enhanced objectless DEM is provided on a computer screen 318 as shown in FIG.

  The execution batch process loops back to block 104 for step (11). Block 104 performs resampling once again on the enhanced, object-free DEM as given in block 110. As before, this resampling is performed at a lower resolution, i.e. a resolution of 1 to 30 meters. The result is given on a computer screen 320 as shown in FIG.

  In block 106, a null fill is performed once again corresponding to step (12). This third null fill is performed on the resampled object-free DEM as provided by block 104 to produce an enhanced land only DEM as provided in computer screen 322 of FIG. This DEM is also called the final land only DEM.

  In block 108, a third DEM subtraction is performed corresponding to step (13). The enhanced land only DEM from step (12) is subtracted from the original DEM in step (1) to generate a DEM only for further improved objects. Only further enhanced objects DEMs are provided on a computer screen 324 as shown in FIG.

  In block 114, corresponding to step (14), a null extension is performed on the DEM only for the further enhanced object. In block 116, null filling is performed corresponding to step (15). Steps (14) and (15) are performed to remove noise from around the object and generate a DEM only for the improved object. The process loops back to block 114 and steps (16) and (17) are performed. That is, once again null extension and null fill are performed to generate a DEM only for the second, further enhanced object as given in the computer screen 326 of FIG.

  At block 118, DEM subtraction is performed. The second improved object only DEM from step (17) is subtracted from the DEM only the further improved object from step (15) to generate a noise only DEM. The noise only DEM is provided on a computer screen 328 as shown in FIG.

  At block 120, another DEM removal is performed (step 19). The noise only DEM from step (18) is only removed from the DEM from the second improved object from step (15) to produce the final object only DEM. Only the final object DEM is given on the computer screen 330 as shown in FIG.

  The final land-only DEM and final object-only DEM were generated as discussed above for steps (2)-(19). Compared to prior art DEMs, these DEMs are improved as a result of the loop iterations performed in steps (2)-(19). Yet another aspect based on the invention that will be discussed is the separation of the final object-only DEM into a building DEM and a vegetation DEM. In other words, only the final object DEM is separated into two separate DEMs, and each DEM can be processed separately.

  The separation of the final object-only DEM into building DEM and vegetation DEM will now be discussed with reference to the flow diagram shown in FIG. 26 and FIGS. Referring initially to the computer screen 200 shown in FIG. 3, the user selects “Separate Building and Tree” in field 227. Then, a “separate building and tree” computer screen 410 as shown in FIG. 27 is displayed.

  Still referring to FIG. 27, the user has the option to select several threshold parameters. Separation is performed on the basis of calculating perimeter per area and height of each object for each object, so a corresponding comparison threshold is set via computer screen 410. For example, a perimeter threshold for the area is set in the field 412. A minimum size for each object to be considered is set in field 414. This field is labeled minimum posts. A chord residue is selected in field 416, which is the width or length of the object being considered. The maximum tree height is selected in field 418. A second threshold value associated with a second per-area test is selected in field 420. This second threshold selected in field 420 may be different from the first threshold selected in field 412.

  To initiate separation of the final object-only DEMs into building DEMs and vegetation DEMs, a perimeter parameter for each object in the final object-only DEM is determined at block 502. For purposes of illustrating the present invention, referring also to FIG. 28, it is given how only the final object DEM is separated into a building DEM and a vegetation DEM. For example, first, a simplified representation representing only the final object DEM as shown in FIG. 25 is provided in frame 600 of FIG.

  As shown in frame 600, the objects include buildings 54 and trees 56 grouped together in the same DEM. The circumference versus area parameter for each object is compared to the threshold selected in block 504 to identify whether each object in the DEM is a building 54 or a vegetation 56. Based on the comparison with the threshold, the object is separated into a building DEM and a vegetation DEM, as shown in frames 602 and 604.

  The data for each object includes a height value, and the height value for each object identified as vegetation 56 in frame 604 is compared to a height threshold at block 506. At block 508, each vegetation 56 in the frame 604 is re-identified as a building 54 if its associated height value is greater than a height threshold. As shown in frames 606 and 608, the building 54 identified as vegetation 56 in frame 604 has been re-identified as a building in frame 606. However, as shown in frame 606, the tall tree 56 in frame 604 has now been identified as a building based on a comparison with a height threshold.

  At block 510, a second perimeter parameter is determined for each object identified as a building at frame 606. Each of the second circumference pair area parameters is compared to a second threshold at block 512. Each building 54 is re-identified as a vegetation 56 at block 514 if the second circumference-to-area parameter is greater than the second threshold.

  At block 516, objects identified as buildings 54 are separated into building DEMs, and objects identified as vegetation are separated into vegetation DEMs. Separate building and vegetation DEMs advantageously allow a more realistic terrain model to be generated with significantly less user intervention. Building DEM is represented by frame 610 and vegetation DEM is represented by frame 612. The method of separating the final object-only DEM into a building DEM and a vegetation DEM ends at block 518. The above steps of separating the final object only DEM into two separate DEMs correspond to step (20) in FIG.

  Here, the remaining steps (21) to (26) will be discussed. Buildings 54 and trees 56 are modeled differently. Block 124 corresponds to step (21) and is optional, but allows the user to manually finish or edit the separation of building 54 and tree 56 if the automated process fails to correctly identify each object.

  At block 126, corresponding to step (22), each vegetation in vegetation DEM 612 is modeled as a 3D point. To convert a point to a list of x (latitude), y (longitude) and z (altitude), the user selects “Generate Point from DEM” in field 228 as shown in FIG.

  Block 128 corresponds to step (23) where the building is modeled as a vector. Modeling buildings as vectors is disclosed in US Pat. No. 6,654,690 discussed in the Background section. At block 130, corresponding to step (24), a texture is placed on the polygon representing the building. In other words, an image is arranged on the polygon, and a realistic appearance is given to the terrain model. RealSite ™ is one commercially available tool that performs this task. RealSite (TM) was developed by Harris Corporation, the assignee of the present invention.

In block 132, SceneBuilder ™ is used to shape all of the geometric shapes and textures generated for display on the computer system 60. SceneBuilder ™ is also a commercially available tool. Using InReality ™ at block 134, a final terrain model for display as shown in FIG. 29 is provided. InReality (TM) is another commercial tool developed by Harris Corporation that allows users to navigate through virtual scenes and perform various analyses. InReality (TM) is designed as a partner to RealSite (TM) software. The process ends at block 136.

FIG. 3 is a block schematic diagram of collecting terrain data and generating a terrain model from the collected terrain data according to the present invention. 3 is a flowchart for generating a terrain model according to the present invention. FIG. 6 is a computer screen display corresponding to the generation of an original DEM according to the present invention. FIG. 6 is a computer screen display corresponding to the generation of an original DEM according to the present invention. FIG. 6 is a computer screen display corresponding to the generation of an original DEM according to the present invention. FIG. 5 is a computer screen display corresponding to resampling of the original DEM according to the present invention. FIG. 5 is a computer screen display corresponding to resampling of the original DEM according to the present invention. FIG. 4 is an illustration of a computer screen display corresponding to a null filling process performed on a resampled DEM according to the present invention. FIG. 4 is an illustration of a computer screen display corresponding to a null filling process performed on a resampled DEM according to the present invention. FIG. 4 is an illustration of a computer screen display corresponding to a null filling process performed on a resampled DEM according to the present invention. FIG. 6 is a diagram of a computer screen display corresponding to an object-only DEM and a DEM subtraction for generating a DEM without the object according to the present invention. FIG. 6 is a diagram of a computer screen display corresponding to an object-only DEM and a DEM subtraction for generating a DEM without the object according to the present invention. FIG. 6 is a diagram of a computer screen display corresponding to an object-only DEM and a DEM subtraction for generating a DEM without the object according to the present invention. FIG. 14 is a computer screen display corresponding to the null expansion process performed for the DEM without an object given in FIG. 13. FIG. 14 is a computer screen display corresponding to the null expansion process performed for the DEM without an object given in FIG. 13. FIG. 4 is a diagram of a computer screen display corresponding to land-only DEM generation according to the present invention. FIG. 4 is a diagram of a computer screen display corresponding to land-only DEM generation according to the present invention. FIG. 5 is a computer screen display corresponding to DEM subtraction to generate an improved object only DEM and an improved DEM without the object in accordance with the present invention. FIG. 5 is a computer screen display corresponding to DEM subtraction to generate an improved object only DEM and an improved DEM without the object in accordance with the present invention. FIG. 6 is an illustration of a computer screen display corresponding to generation of a final land only DEM and further enhanced object only DEM in accordance with the present invention. FIG. 6 is an illustration of a computer screen display corresponding to generation of a final land only DEM and further enhanced object only DEM in accordance with the present invention. FIG. 6 is an illustration of a computer screen display corresponding to generation of a final land only DEM and further enhanced object only DEM in accordance with the present invention. FIG. 4 is a diagram of a computer screen display corresponding to generating a second DEM only second improved object based on null fill / null extension in accordance with the present invention. FIG. 4 is a diagram of a computer screen display corresponding to noise-only DEM generation in accordance with the present invention. FIG. 6 is a computer screen display corresponding to the generation of a DEM only for the final object according to the present invention. FIG. 26 is a flow chart for separating only the final object given in FIG. 25 into a building DEM and a vegetation DEM. FIG. 7 is a computer screen display for setting parameters related to separation of a final object-only DEM into a building DEM and a vegetation DEM, in accordance with the present invention. FIG. 4 is a diagrammatic representation of steps for separation of a final object-only DEM into a building DEM and a vegetation DEM, in accordance with the present invention. It is a figure of the computer screen display of the topographic model produced | generated based on this invention.

Claims (10)

A computer-implemented method for processing a digital elevation model (DEM) containing data about multiple objects:
Determining a circumference-to-area parameter for each object in the DEM;
Comparing the circumference-to-area parameter for each object with a threshold to identify whether each object in the DEM is a building or a vegetation. The computer-implemented method of claim 1, wherein the data for each object includes a height value, further comprising:
Comparing the height value for each object identified as vegetation with a height threshold;
Re-identifying each vegetation as a building if the height value associated therewith is greater than a height threshold. The computer-implemented method of claim 2, further comprising:
Determining a second circumference-to-area parameter for each object identified as a building;
Comparing each second perimeter parameter to a second threshold;
Re-identifying each building as vegetation if the second circumference-to-area parameter is greater than a second threshold.   The computer-implemented method of claim 1, wherein objects identified as buildings are separated into building DEMs, and objects identified as vegetation are separated into vegetation DEMs. Modeling the building in the building DEM;
Modeling the vegetation in the vegetation DEM;
The computer-implemented method of claim 4, further comprising displaying the modeled building and vegetation on a display. A computer system for terrain modeling:
A processor for processing a digital elevation model (DEM) containing data for a plurality of objects, said processing comprising:
Determining a circumference-to-area parameter for each object in the DEM;
Comparing the perimeter parameter for each object to a threshold to identify whether each object in the DEM is a building or vegetation;
The computer system further comprising a display coupled to the processor for displaying a terrain model based on the processing. The computer system of claim 6, wherein the data for each object includes a height value, wherein the processor:
Compare its height value for each object identified as vegetation with a height threshold;
Re-identify each vegetation as a building if its associated height value is greater than the height threshold;
Computer system. 8. The computer system of claim 7, wherein the processor is:
Determining a second circumference-to-area parameter for each object identified as a building;
Comparing each second perimeter parameter to a second threshold;
A computer system that re-identifies each building as a vegetation if the second perimeter parameter is greater than a second threshold.   The computer system of claim 6, wherein an object identified as a building by the processor is separated into a building DEM, and an object identified as vegetation by the processor is separated into a vegetation DEM.   The computer system of claim 9, wherein the processor models a building in a building DEM and models a vegetation in a vegetation DEM to define a terrain model displayed on the display.

Images